MRI Eye Shape Linked to Faster High Myopia Progression
Key Takeaways
- Highly myopic eyes in this long-term Chinese cohort continued to show annual myopic shift and axial elongation over follow-up.
- Eyes classified as deformed on 3D magnetic resonance imaging (MRI) had significantly faster spherical equivalent and axial length progression than non-deformed eyes.
- MRI-based globe phenotyping separated six ocular shape categories, with spheroidal and ellipsoidal eyes grouped as non-deformed and conical, nasally distorted, temporally distorted, and barrel-shaped eyes grouped as deformed.
- In exploratory modeling, random forest had the lowest error for annual axial length progression, while XGBoost had the lowest error for annual spherical equivalent progression.
Investigators analyzed a prospective cohort within the Zhongshan Ophthalmic Center-Brien Holden Vision Institute High Myopia Cohort Study in this six-year study. Eligible participants were adults aged 18 to 70 years with bilateral high myopia defined by spherical power less than −6.00 diopters. The analysis included 150 eyes from 75 participants, with mean follow-up of 6.70 ± 2.66 years and visits at 2-year intervals. Baseline assessment combined high-resolution 3D MRI, ocular biometry, and cycloplegic refraction, with spherical equivalent (SE) and axial length (AL) tracked longitudinally. Ocular shape was classified as spheroidal, ellipsoidal, conical, nasally distorted, temporally distorted, or barrel-shaped, then grouped as non-deformed for spheroidal and ellipsoidal eyes and deformed for the remaining four phenotypes; both eyes were included in analysis with mixed-effects modeling used to account for longitudinal change.
Across the cohort, annual progression averaged −0.36 D/year for SE and 0.07 mm/year for AL. Deformed eyes progressed faster than non-deformed eyes, with descriptive rates of −0.49 versus −0.27 D/year for SE and 0.09 versus 0.05 mm/year for AL. Progression was seen across ocular shape subtypes, although phenotype-specific differences were reported descriptively rather than as formal subgroup comparisons.
In adjusted mixed-effects and machine-learning analyses of ocular shape progression, baseline deformity remained associated with faster SE progression (β = −0.20; 95% CI, −0.30 to −0.11; P < 0.001) and greater AL elongation (β = 0.05; 95% CI, 0.02–0.07; P < 0.001). As a secondary exploratory finding, random forest had the lowest RMSE for annual AL progression at 0.067, while XGBoost had the lowest RMSE for annual SE progression at 0.644. The modeling results were secondary to the main longitudinal finding that deformed baseline globe shape tracked with faster progression.
The cohort was observational, MRI was performed only at baseline, and some phenotype subgroups were small, especially temporally distorted and barrel-shaped eyes. MRI-derived contours reflected intraocular fluid shape rather than direct scleral contours, and the study did not include hyperopic, emmetropic, or mild-to-moderate myopia control groups. Because the cohort was drawn from highly myopic adults in China, the findings are most directly applicable to a Chinese high-myopia population rather than automatically representative of U.S. patients. The authors suggested that baseline MRI-based ocular shape may help identify highly myopic eyes at higher risk of faster biometric progression, while the prediction models showed limited individualized accuracy.
The authors concluded that biometric progression persisted during long-term follow-up in highly myopic eyes and was faster when baseline globe shape was deformed rather than non-deformed. They described MRI-based ocular phenotyping as a possible observational tool for risk stratification within this cohort rather than a practice directive.
Clinician Questions
How did the investigators define deformed versus non-deformed ocular shape in high myopia?
The cohort used six MRI-defined phenotypes: spheroidal, ellipsoidal, conical, nasally distorted, temporally distorted, and barrel-shaped. Spheroidal and ellipsoidal eyes were categorized as non-deformed, whereas conical, nasally distorted, temporally distorted, and barrel-shaped eyes were categorized as deformed.
Which highly myopic eyes were included in this MRI cohort, and what does that mean for applicability?
The cohort included adults aged 18 to 70 years with bilateral high myopia defined as spherical power less than −6.00 D, recruited within a Chinese cohort. The study did not include hyperopic, emmetropic, or mild-to-moderate myopic control eyes, so the findings apply to highly myopic eyes rather than the full refractive spectrum.
Why were phenotype-specific progression rates treated as preliminary in this high myopia analysis?
Those rates were descriptive observations rather than formal subgroup comparisons, and some individual shape groups were small, especially temporally distorted and barrel-shaped eyes. That makes the phenotype-level patterns best interpreted as early signals within this cohort rather than definitive comparisons across all six MRI-defined shapes.
Recommended Reading
- For more on axial elongation in myopia: Racket Sports Rank Highest for Pediatric Myopia Control